d@0: d@0: Next: Guru Complex DFTs, d@0: Previous: Interleaved and split arrays, d@0: Up: Guru Interface d@0:
The guru interface introduces one basic new data structure,
d@0: fftw_iodim, that is used to specify sizes and strides for
d@0: multi-dimensional transforms and vectors:
d@0:
d@0:
typedef struct {
d@0: int n;
d@0: int is;
d@0: int os;
d@0: } fftw_iodim;
d@0:
d@0:
d@0: Here, n is the size of the dimension, and is and os
d@0: are the strides of that dimension for the input and output arrays. (The
d@0: stride is the separation of consecutive elements along this dimension.)
d@0:
d@0:
The meaning of the stride parameter depends on the type of the array
d@0: that the stride refers to. If the array is interleaved complex,
d@0: strides are expressed in units of complex numbers
d@0: (fftw_complex). If the array is split complex or real, strides
d@0: are expressed in units of real numbers (double). This
d@0: convention is consistent with the usual pointer arithmetic in the C
d@0: language. An interleaved array is denoted by a pointer p to
d@0: fftw_complex, so that p+1 points to the next complex
d@0: number. Split arrays are denoted by pointers to double, in
d@0: which case pointer arithmetic operates in units of
d@0: sizeof(double).
d@0:
d@0: The guru planner interfaces all take a (rank, dims[rank])
d@0: pair describing the transform size, and a (howmany_rank,
d@0: howmany_dims[howmany_rank]) pair describing the “vector” size (a
d@0: multi-dimensional loop of transforms to perform), where dims and
d@0: howmany_dims are arrays of fftw_iodim.
d@0:
d@0:
For example, the howmany parameter in the advanced complex-DFT
d@0: interface corresponds to howmany_rank = 1,
d@0: howmany_dims[0].n = howmany, howmany_dims[0].is =
d@0: idist, and howmany_dims[0].os = odist.
d@0: (To compute a single transform, you can just use howmany_rank = 0.)
d@0:
d@0:
A row-major multidimensional array with dimensions n[rank]
d@0: (see Row-major Format) corresponds to dims[i].n =
d@0: n[i] and the recurrence dims[i].is = n[i+1] *
d@0: dims[i+1].is (similarly for os). The stride of the last
d@0: (i=rank-1) dimension is the overall stride of the array.
d@0: e.g. to be equivalent to the advanced complex-DFT interface, you would
d@0: have dims[rank-1].is = istride and
d@0: dims[rank-1].os = ostride.
d@0:
d@0: In general, we only guarantee FFTW to return a non-NULL plan if
d@0: the vector and transform dimensions correspond to a set of distinct
d@0: indices, and for in-place transforms the input/output strides should
d@0: be the same.
d@0:
d@0:
d@0:
d@0: